How should a reseller package releases so customers can manage later changes?
Release manifests and stable identifiers let resellers track schema changes, corrected labels and withdrawn cases across customer downloads without silently altering evaluation inputs.
Ask for a release manifest containing `package_id`, `release_version`, `schema_version`, `case_ids`, `file_checksums` and `licence_reference`. The licence reference should identify the applicable document; it should not be treated as a machine-readable grant of unspecified rights.
Reseller transformations can have their own manifest with `upstream_release_version`, `transformation_version` and a mapping from upstream case identifiers to customer-facing identifiers. This preserves traceability when a marketplace changes directory layouts or converts JSONL into Parquet.
Agree on how updates would be represented. A corrected evaluation label, a schema change and a removed case have different consequences for customers. An update record can state `change_type`, `affected_case_ids`, `replacement_refs` and `evaluation_impact`, allowing customers to decide whether prior runs need review.
During a pilot, test that a customer can identify an installed package and compare it with a proposed replacement. Keep versions immutable within this packaging design, rather than silently replacing files under an existing version. This is a requested integration approach, not an assumed commitment to any update schedule or support level.
Establish update mechanics during the pilot so downstream customers can identify exactly which generated material they are using.
- Audience:
- Licensed resellers
- Content type:
- Integration
- Buyer question:
- Scope and getting started
Content release: r-a220eb6e099b243b
Related questions
- How can a pilot cover closed accounting periods without modelling an entire finance function?
- What licence questions should we resolve before using generated material in an agent pipeline?
- What should a correction pilot brief specify before generation begins?
- What integration details should we settle before loading a pilot into our agent harness?
- What is a useful first scope for an agent correction pilot?
- How can a pilot expose a scope boundary without implementing an entire finance system?